allenjack / gate Goto Github PK

Great work and implement，can you share the processed movilen dataset or process script?
Thanks!

My problem may be too simple, but it confuses me very much. Please pay attention to it~~

In auto encoder, I know, it's trained x to x~, so it can learn more accurate hidden layer. My question is: when we make predictions, test data is items which users don't rate. So, what's the input? where's input in the network architecture? is gating layer's input? or Encoder_r's input? or other?

Thanks for your attention!

Hi allenjack! I just tried to run your code in the exact way you showed in readme file. But I got error information as following:

DEBUG:main:epoch:0
Traceback (most recent call last):
File "run.py", line 251, in
main()
File "run.py", line 247, in main
train_model(train_matrix, item_content_matrix, item_relation_matrix, word_seq, test_set, args)
File "run.py", line 204, in train_model
batch_x_weight = Variable(torch.from_numpy(batch_x_weight).type(dtype), requires_grad=False)
TypeError: can't convert np.ndarray of type numpy.int32. The only supported types are: double, float, float16, int64, int32, and uint8.

Thank you for your attention!